• Non ci sono risultati.

A gamification framework demonstrating a complete cycle of vehicle driver performance evaluation

N/A
N/A
Protected

Academic year: 2021

Condividi "A gamification framework demonstrating a complete cycle of vehicle driver performance evaluation"

Copied!
111
0
0

Testo completo

(1)

GAUTAM RAVINDRA DANGE 1

University of Genoa

Department of Naval, Electrical, Electronic and Telecommunications Engineering

Gautam Ravindra Dange

A Gamification framework demonstrating a

Complete Cycle of Vehicle Driver

Performance Evaluation

Ph.D. Thesis

Doctoral Course in Interactive Cognitive Environments

Cycle XXX

Supervisor: Dr. Francesco Bellotti

(2)

GAUTAM RAVINDRA DANGE 2

Contents

Abstract ... 4

Acknowledgements ... 6

Disclaimer ... 7

Acronyms and abbreviations ... 8

1. Introduction ... 9

2. State-of-the-Art ... 13

3. Assumptions and Scope ... 20

4. Architecture ... 21

4.1 Social Gaming prototype scenario ... 24

4.2 Competitions: ... 25

4.3 SG-CB (Serious Games & Community Building) Application ... 25

4.4 Virtual Coins ... 26

5. Vehicle Performance Evaluation ... 28

5.1 Vehicle Data Analysis: ... 29

5.1.1. Basis of Evaluation: ... 33

5.1.2 Normalizations ... 33

5.1.3 Observations, Possible False conceptions and clarifications: ... 36

5.2 Signal Correlations: ... 37

5.3 Penalty Evaluation ... 42

5.4 Evaluation of Vehicle Driver Performance ... 48

5.4.1 Time of Evaluation ... 48

5.4.2 Evaluation Configuration Parameter Tuning: ... 49

5.4.3. Competition Instance and Participation of Users ... 49

6 Evaluators ... 51

6.1 Linear Distance ... 51

6.2 K-NN (Nearest Neighbours) ... 53

6.3 Dynamic Sliding Window ... 55

6.4 Fuzzy system based Dynamic Sliding Window (FDSW) ... 57

Comparisons ... 60

6.5 In App Evaluation Implementation using FDSW ... 61

7. Gamification and Performance Visualization ... 64

7.1 SG-CB Application ... 64

(3)

GAUTAM RAVINDRA DANGE 3

7.3 Possible Achievements ... 67

7.3.1. Awareness and Coaching: ... 67

7.3.2 Motivation: ... 68

7.3 Types of Gamification Mechanics: ... 68

(i) Score/ points ... 68

(ii) Virtual Coins ... 68

(iii) Awards ... 68

(iv) Notifications... 68

(v) Leader-boards ... 68

(vi) Events visualization... 69

(vii) Progress Tracking ... 69

(viii) Games with level and Unlocking new levels ... 69

(ix) Game with randomness variant ... 69

7.4 Gamifications and its impact on user ... 71

8. The Cloud and Framework ... 72

8.1 Server Modules and Services ... 74

8.2 Staged Competition ... 75

8.3 Staged Competition Manager (Smart Phone Application) ... 82

9. Experiments ... 91

9.1 Simulations: ... 92

9.1.1 Implementation Overview ... 92

9.1.2 Aggregation Server ... 93

9.1.3 Driver Performance Evaluator: ... 94

9.1.4 Vehicle Simulation Unit: ... 95

9.2 Real-Time On-Site Tests ... 95

9.2.1 Test Site description 1 Goteborg, Sweden ... 95

9.2.2 Test Site description 2 Turin, Italy Eco Challenge 2017, Feb 2016 ... 97

9.2.3 Test Site description 3 Trento, Italy [Remote Test, from Genova] ... 98

9.2.4 On-Site Test with Smart-phone based (In-App) Evaluation: ... 99

9.2.5 Test Site description 4 Genova, Italy, [Smart Phone Evaluator only] ... 100

10 Conclusion Limitations and Future work ... 101

(4)

GAUTAM RAVINDRA DANGE 4

Abstract

Training through a gamified environment motivates the users in achieving optimal outcome and reduces the complexity of learning by adding factor of entertainment in it. The deployment of serious games in automotive industry is a major leap in technological grounds, as it’s a best way to inculcate safe driving patterns to reduce the fatalities and enhance resource usage which includes car accessories and fuel. The Ph.D. thesis represents Gamification platform aimed to Green Mobility and Safe Driving. Good driving behaviour is a significant factor for road safety and green mobility. Thus, major tasks involve the evaluation of vehicle signals for analysing driver behaviour. How this infrastructure analyses vehicle driver performance, how information gets stored, processed and then further displayed graphically, statistically or in form of gamification, can be found in this thesis.

The Framework consist of 3 tier architecture, which includes Evaluators, Cloud running RESTful API services and Gamification Applications. Evaluators are either running inside Car or in a driver's smartphone. The Evaluators which are deployed inside car, called In-car Evaluators, do run in OSGi (Open System Gateway Initiative) Environment. The Evaluators which are run as Smartphone Application, called as In-Application evaluators. The In-Car evaluators use the raw vehicle signals for driver evaluation, whereas the In-Application evaluators use Mobile Sensors like Gyroscope, Accelerometer and GPS Data for evaluating a vehicle driver. For In-car evaluation the vehicle comprises of an OSGi environment, where the evaluators and signal receivers are deployed in the OSGi environment as Jar bundles. Vehicle Signals are transmitted from CAN bus to Vehicle Data provider bundles then to vehicle Data Consumer. The Different algorithms uses the signals from Vehicle data consumer for Driver Evaluation. The Scores, Events, time stamps and GPS latitude longitude information is sent to Cloud Server. The driver get evaluated continuously and partial results submitted to cloud server one every 20 or 30 seconds. The driver gets compared with peers, in context of competition, the driver who gets highest scores, gets ranked at the top, and gets incentives.

The Cloud server is a Tomcat Server deployed on google cloud. The MySQL database have relations which stores information about Users, Evaluators, Games, Application, Competitions, Competition Subscribers, Awards, Milestones, time-slots, Scores and Events. The Evaluators and Gamification Applications uses RESTful APIs for storing, querying, updating or removing in required contents. The Competition Manager tool is developed which allowed different Geographical and time-slots based competition. Once the vehicle driver subscribes for

(5)

GAUTAM RAVINDRA DANGE 5

the competition he/she automatically participate in the competition whenever he/she arrives at mentioned area in mentioned time-slots. The Gamification Application includes the application for checking out scores, Events and Incentives gained by the user. The Architecture provides and encourages implementation of serious games on the basis of Driver Performance. Using exposed RESTful APIs, game designer can access the user and his scores information on regular interval and can use this scores as an input to his/her game logic. This activity research activity was carried as a part of SG-CB (Serious Games and Community Building) module of TEAM (Tomorrow’s Elastic Adaptive Mobility) project but had few different goals in parallel. Being in TEAM project thesis mentioned application, algorithms and RESTful API implementations had got tested on Car models which includes BMW sedan and Mercedes-Benz (s-class) & FIAT 500 at 2 different sites which includes ASTA (Active Safety Test Area), Goteborg, Sweden and Centro Ricerche Fiat (CRF) Turin, Italy. This project have got tested successfully under TEAL IP-7 and infrastructure encourages the games designer to use Serious Data for Gamification by providing Serious Gaming implementation platform.

(6)

GAUTAM RAVINDRA DANGE 6

Acknowledgements

It would not have been possible to write this doctoral thesis without the help and support of the kind people around me, to only some of whom it is possible to give particular mention here.

My parents have given me their explicit support throughout, as always, for which my mere expression of thanks likewise does not suffice.

This thesis would not have been possible without the help, support and patience of my supervisor, Dr. Francesco Bellotti, not to mention his advice and important critics, which kept me doing better and initiative.

I would like to warmly thank my colleagues in laboratory Pratheep Kumar Paranthaman and Nikesh Bajaj for their friendship and advices whenever required.

The good advice and support of Dr. Riccardo Berta and Alessandro de Gloria, Laboratory at DITEN, have been invaluable, for which I am extremely grateful.

I am grateful for the support from different partners of TEAM IP 7 project, includes Centro Ricerche FIAT Trento and Turin, Italy Teams, Fraunhofer Institute for Open Communication Systems FOKUS, Berlin, Germany, Volvo Group Trucks Technology (GTT), Goteborg, Sweden and Objective Software consulting solutions, Turin, Italy.

The TEAM IP 7 builds a big interactive partnership and it is a completely heterogeneous, very informal, but wonderful, group of colleagues and friends, with multiple intersections between the people from different backgrounds, includes experts of vehicle dynamics, chips designers, and experts in city traffic control.

(7)

GAUTAM RAVINDRA DANGE 7

Disclaimer

I, Gautam Ravindra Dange (“the Author”), wrote this doctoral thesis in my personal capacity while studying in University of Genova, as Doctorate student.

This thesis has been written using open information sources only, i.e. public domain information gathered from unclassified knowledge sources, being the majority of them available in the internet, and not subject to any disclosure restriction of any kind. This includes some of the yet published results of various research activities carried out in TEAM IP 7 project and own experimentations as a part of research and development.

All cited sources other than self-cited publications (texts, patents, data and images) have been duly referenced, where possible and where required. All cited trademarks and patents have been indicated in the property of their respective owners, where possible and where required.

The Author has made reasonable efforts to ensure that all information and material included in this thesis was accurate at the time of writing.

The Author has not received any remuneration, consideration or benefit for mentioning, reporting, including, or commenting about any paper, program, product, policy, law, theory, hypothesis, or intellectual property. For any errors or inadequacies that still may remain into this work, of course, the responsibility is entirely deemed on my own

(8)

GAUTAM RAVINDRA DANGE 8

Acronyms and abbreviations

TEAM : Tomorrow’s Elastic Adaptive Mobility (European IP 7 Project) SG-SB : Serious Games and Community Building

FOKUS : Fraunhofer Institute for Open Communication Systems

CRF :

Centro Ricerche Fiat

HTTP : Hyper Text Transmission Protocol OSGi : Open System Gateway Initiative

XML : eXtensible Markup Language

JSON : JavaScript Object Notation

JAR : Java ARchive

WAR : Web application ARchive

VC : Virtual Coins

VDP : Vehicle Data Provider (Service Bundle) VDC : Vehicle Data Consumer (bundle)

PE : Penalty Event

REST : Representational state transfer K-NN : K-nearest-Neighbour (Algorithm)

DSW : Dynamic Sliding Window (Evaluation Algorithm)

CAN : Controller Area Network (Vehicle supported bus for signal expose) RPM : Rotations per minute (Engine load detection)

API : Application Programming Interface GPS : Global Positioning System

(9)

GAUTAM RAVINDRA DANGE 9

1.

Introduction

It is always the driver behaviour that has the direct impact on all the factors that are concerned to efficient driving characteristics such as the road safety, fuel efficacy or the reasonable driving patterns. Also let it be the fatalities or vehicle collisions, it is again a genesis from the behaviour of the driver [9] [10]. As the driver behaviour holds a significant role in modelling the safe and green driving patterns, there are many models that are emerging under this roof. Providing necessary coaching to drivers under simulated environments would act as a test bed to facilitate the process of designing various safe driving methodologies [11]. Other approaches would be the prior and post analysis of driver behaviour with reference to the impact [12] [13], by predicting the outcomes and adjusting the current behaviour. However there is a mild patch between the simulation and real-world entity, most of the simulated results have to be adjusted to cope with the real-world situations. These safety and efficient driving measures must be approached from the perspective of system design [10], where the modelling must accommodate the dynamic environment. The major aim of my research activity is to inculcate the better driving practices using driver behaviour evaluation and social gaming scenario. This research activity was carried as a part of TEAM (Tomorrow’s Elastic Adaptive Mobility) project [14].

For an early evaluation, data from vehicle signals collected in a test run conducted by Centro Ricerche Fiat (CRF) [15] in an area in the surroundings of Trento, Italy was used. As a viewpoint of my approach, the competitive gaming scenario would induce a good conduct in driving pattern and the evaluators used would ensure that the proper driving behaviour is exhibited. As the evaluators are comprised of four different approaches such as Linear distance to categorize the harsh and smooth driving patterns, and K-Nearest Neighbour approach to plot the relativity between the vehicle signals and to penalize the harsh behaviours attempted by the drivers whereas Dynamic Sliding Window captures the deviation from optimal level and track the events and Fuzzy based Dynamic sliding window to support same algorithm for different platforms without any change in code. Therefore the evaluators are designed to analyse various driving styles and fit to the context of the dynamic environment.

Most of the existing models in serious gaming context for automotive industry aims at improvising driver behaviour by various methods. But, there is another viewpoint for this approach (i.e.) out of many methods that could enhance driver behaviour, the chief methodology would be the qualitative driver performance analysis. I believe that, the first step

(10)

GAUTAM RAVINDRA DANGE 10

in attaining better driver behaviour would be devising a qualitative evaluation pattern that would analyses and provide a detailed statistical report of driver performance. These analysis can be of great assistance in comparing the individual driver performance with the performance of peers, also the comparative analysis can be used for conducting serious games to inculcate the efficient driving behaviour. Having all these considerations as foundation, development of a comparative performance assessment schema that would analyse the driver behaviour on absolute and social comparison basis is noteworthy.

The connectivity and availability of transportation systems are increasing, and this enables the traveller to navigate to any destination comfortably. The types of transportation services are extended by more modern approaches like car sharing or rides. But although there are many possible ways to reach a destination, own transport is the preferred mode, because of comfort-ability and easiness. The use of the own car increases number of vehicles on road, which in turn contributes traffic congestion, air pollution and accidents. To avoid the increase of air pollution and critical accidents various solutions can be applied. One of these solutions is directly provided by car manufacturers. They aim to reduce fuel consumption, emission of cars, and increase of safety. One primary factor in reducing the air pollution and improving the safety is by altering the behaviour of the driver. If the driver is aware and willing to be follow better driving practices, the exhaust of pollution is low and safety is high. Otherwise if the driver is driving harsh or negligent, it will cause a high exhaust of pollution and safety risk for him/her as well as others. If one could influence the driving habits and awareness of a driver in an appropriate way, there will be a higher safety and a lower air pollution. Thus introducing an approach that will help individuals to drive in an appropriate way, e.g. to lower the air pollution and the risk of accidents. It is essential to convince the driver to stick to a better driving behaviour. The use of serious games approach can captivate the drivers to maintain optimal driving behaviour and by this methodology the drivers directly benefit by earning rewards. Every time the user showcase better driving behaviour, he/she will earn points and these points gets transferred as virtual coins to the user profile. The users can spend virtual coins on real-world applications (bus tickets, parking tickets, etc.). In general, the gaming approach may convince people to drive in an optimal way even if they would not consider about air pollution or safety issues.

The research has main objective of evaluating driver performance. Harsh or rash instances are fraction of single percent of whole driving, so, evaluation need to be penalty based. In the real time competition, whenever driver apply any harsh pattern, the algorithm detects

(11)

GAUTAM RAVINDRA DANGE 11

this moves and assign a Penalty, for this pattern. This penalty will inversely affect his/her competition score. If Harshness of driving is significantly high, and driver get high penalty, then driver need to be notified about this behaviour with the details about date, time, harshness pattern and location where this harsh behaviours are observed, and few instant feedbacks also can be given to driver by acoustic notes. This process would be helpful making driver aware of his/her mistakes and reducing negligence.

The thesis explains the serious game building framework and narrates one demonstration of implemented example of a transportation scenario for vehicle driver performance evaluation. Firstly the explanation about Vehicle driver performance evaluation framework using RESTful APIs and Gamifications will be explained and then the idea of transparent evaluation and gamification methodology will be found in this thesis.

Project research activity of the PhD is about providing a gamification framework for game designers by providing a platform with which different game designers can develop serious games, without worrying about for which domain the serious game is created. The framework clearly distinguish between the serious information provider units with gamification unit which user this information. So, without the knowledge of real world understanding of the problem game designer can focus on score and period at which they with receive the score in numbers. Basically game designer will provide a username to the cloud with the frequency at which game score is expected, the cloud server will responds with the username and his performance considering the frequency of evaluation. Gamification can ask scores on a frequency which is lower than a frequency of evaluation, that is, if evaluation is sending data on every minute then game can’t ask for 30 second interval, but can ask score on any other interval greater than 1 minute. Scores data not only gets stored on server as raw data, but server has different data structures and functions which calculates the score of hour using data of minute samples, and calculated daily data by calculating hourly samples in server in different data structures. This process allow different gamifications to use different periodic intervals to fetch the user data. For an example there can be a game called “Talking Tom”. So Tom (Cartoon cat) game ask the server data of whole day performance of the vehicle user U, and with that score, Talking Tom shows different faces happy face, sad face or angry face. If whole day a drivers drives very optimal then Talking Tom can show his happy face on his game scene, or if driver applies some harsh brakes then at the end of the day Talking Tom would show angry face as a question on his driving behaviour. So basically game designers can implement any game with this data, by requesting server for the information which is in fact a serious information. At the

(12)

GAUTAM RAVINDRA DANGE 12

same time, there can even more evaluators can be added for different comparative reason like person diet values. The connected framework would bring evaluation and gamification together using RESTful API calls, and framework will be responsible of providing evaluations related descriptions of period, score level, optimal score level and conversion of scales for different games according to their requirement.

Chapter 2 will discuss State of art in the domain of vehicle driver evaluation and serious games. Chapter 4 will discuss 3 tier Architecture of the proposed solution. Chapter 5 will discuss how vehicle driver performance is getting evaluated and its different types. Chapter 6 will discuss various sources which are used to get raw vehicle signals and which evaluators used to evaluate the performance. Chapter 7 will discuss about gamifications and performance visualizations for infotainment and serious games. Chapter 8 will discuss about a model, which automate the configurations required for vehicle driver evaluation. Also discuss about a platform which create a transparency between evaluators and games. This provides a facility by which game designer can create Serious game without knowing details of serious evaluation details given by cloud. Chapter 9 discuss about the experiments performed in Goteborg, Sweden and Turin, Italy. Chapter 10 discuss conclusions and possible future work in the domain respected to mentioned research.

(13)

GAUTAM RAVINDRA DANGE 13

2.

State-of-the-Art

When designing a driver behaviour model there are many aspects that are taken into account such as data acquisition tools equipped with the system, extraction of vehicle signals based on the dynamics and most importantly the evaluation methods (algorithms) to process and analyse the vehicle signals to estimate the performance of the driver. To estimate the performance of the driver, [1] conducted an experiment involving in acquiring the time-series steering angle data during the lane changes by the driver. The experimental setup was under a controlled simulated environment and driver behaviour model was designed using conditional Gaussian model on Bayesian network and there was also a comparison along with traditional Hidden Markov Model. The final reports stated that, there were some minor problems in predicting the sudden changes in the environment, such as drastic change of steering wheel for dynamic happenings. Another comparison approach performed by [2] involved the estimation of driver behaviour model by implementing Gaussian mixture model and piecewise auto regressive exogenous (PWARX). The major task was the car following approach, where both of the probabilistic models outperformed each other on various occasions, for instance PWARX produced a good prediction in handling the Gas and brake pedal signals compared to Gaussian mixture model. However both of the algorithms performed well for the limited signals, as in the main focus was on gas and brake pedals only. The different approach that was implemented in car following task using Dirichlet Process Mixture [3]. The outcome stressed on the fact that, the metrics observed from this experiment can be a good step for future analysis in developing context adaptive system. These are some of the significant probabilistic models on current practice and let’s flip the side of the approach and look out the implementation of driver behaviour models using machine learning and adaptive algorithms. The ultimate challenge comes only when things are tested on the real-world applications in a rapidly changing environment, on such aspect the experiment [4] carried out by research team, comprised of “UYANIK”, a passenger car equipped with multiple sensors and CAN architecture for data acquisition was put up on real–world testing in estimation of driver behaviour. This was also a collaborative research conducted by NEDO (Japan), Drive-safe Consortium (Turkey) and the Trans-European Motorway (TEM), Istanbul. Almost the passenger car was equipped with all possible metrics to gather the signal, so as to make the prediction in an enhanced way. Post the data acquisition by CAN-bus, the signals were sent for processing to a SVM (Support vector

(14)

GAUTAM RAVINDRA DANGE 14

machines) model, the labelled data were classified by SVM and the results of SVM were forwarded to a Hidden Markov Model (HMM), which performed the final prediction of driver behaviour. As the processing involved ample number of input signals, the classification and prediction using the machine learning approach was feasible by the system and the outcomes were formulated in an efficient manner. ANN was used as a major tool to classify the driving styles in various roads in the work carried out by [5].

The vehicle signals such as speed, engine RPM and acceleration were extracted and sent to the remote data centre, over there the analysis of data is performed by the neural networks. The ANN would categorize the types of roads (urban, suburban and highway) in which the vehicle was navigating and also the characteristics of the driver as well. However there exists various neural networks architectures, but only Multi-Layer Perceptron (MLP) is preferred predominantly [6]. Different architectures of neural networks must be experimented, so as to emerge out with better solutions for the problems in analysis tasks. On the basis of the state-of-art study on various driver behaviour modelling algorithms, there are many evident facts in handling the approach, firstly the probabilistic and mathematical models were considerably well on simulated and controlled driving environments and they exhibited a certain amount of error ratios and uncertainty when it came to real-world applications[1], whereas on other side the Machine Learning algorithms had the tendency to adapt to sudden changes in the environment and performed well with viable accuracy on all aspects such as classification, estimation and importantly the prediction. Another significant issue that has to be noted with respect to adaptive algorithms is, the incoming data signals should be more in number, because that particular aspect makes the task of evaluation a better one. It is also necessary for the evaluation parameters to hold the ability to adapt to environments, as most of the implementation happens on real-world applications. So it’s always necessary to design the system architecture with reliable model that would facilitate for the varying environments.

The application of Serious games in various domains such as military [29], corporate [30], healthcare [16] and driving [17] have created huge impact in educating and training people though a gamified environment. Indulging in serious games environment can significantly improvise the skills of players such as decision-making, teamwork, leadership and collaboration in a controlled manner. The deployment of serious games in automotive industry is a major leap in technological grounds, as it’s a best way to inculcate safe driving patterns to reduce the fatalities and enhance resource utilities. Some models involving the games for supporting safe driving such as: the experiment conducted by [18] using a motivational method of displaying

(15)

GAUTAM RAVINDRA DANGE 15

avatar on screen, which cheers the driver during good performance. Fiat Eco drive [19] an application promoted eco driving and iCO2 game [20] gathered driver behaviour data from many users for enhancing Eco driving. There were certain research projects centric on human-machine interface (HMI) for enhancement of driving modalities such as: Project COMUNICAR [25] comprised of human-machine interface (HMI) for managing the driver information system ranging from entertainment to environment safety and the Adaptive integrated driver-vehicle interface (AIDE) [24] project which addresses the behavioural and technical issues oriented to adaptive human-machine interface. The “I-GEAR” [21] project (incentives and gaming environments for automobile routing) used serious games to understand the driver behaviour by creating certain scenarios that driver would comply. Whereas, the serious games IC-DEEP was developed to assess driver performance using in-vehicle Information System (IVIS) for tracking the distraction in driving process[22] and IC-DEEP was also used to test an Advanced Driver Assistance Systems(ADAS) to attain cost effectiveness in testing process[23].

The advancement in automotive industry comprises of numerous embedded applications (in-vehicle devices) to facilitate the driving experience as pleasant and safe for the users. The abilities of these in-vehicle devices are limitless and one major factor that has to be addressed over here is, what are the implications of these devices on the users (drivers)? Because, it’s the ability and tendency of the driver that determines the road safety and environmental aspects to a greater extent and also, the inadequate driver performance incurs many fatalities and chaos amongst the road users [31]. So, this factor gives an emergence and a need for the in-vehicle systems (equipped with the user-friendly framework) that will foster the green and safe driving abilities to disarm the on-road hazards. As it’s important for an in-car HMI (Human Machine Interface) to provide the drivers with comprehensive information to aid in safe mobility [32] [33]. The training process will induce the knowledge and awareness to driver about the driving circumstances and for this motive, use of serious games concept and smartphone utilities were perfect match. Serious games are a purpose based tool rather than an aspect of entertainment [34] [35] and they are widely used in various sectors such as education, military, healthcare and business [36] for captivating users to comprehend an information. Compared to traditional games that focus only on the engagement of the users, serious games does have a much higher responsibility, as it has to leverage and instil the concepts as well [37] [38]. An intense care and responsibility are needed in crafting a serious game framework and this leads to a prominent aspect of using instructional design (the use of technology and multimedia tools to enrich the informative content). Projecting a serious game concept through an instructional

(16)

GAUTAM RAVINDRA DANGE 16

design (representation on HMI) promotes the higher degree of visual feedback for the user and this visual feedback takes a substantial effect on driving conditions [39]. Thus, to inculcate safe driving behaviour and establish a collaborative gaming structure, I developed a smartphone-based serious gaming architecture with diversified game logics to enhance the driver performance. The reference architecture is a Service-Oriented architecture [40]. The game logics act as an informative box that supplies the feedback to the users (drivers) in various ways such as scores, virtual coins (incentives that can be used on real-world entities) and live game for displaying the performance evolution. As the serious game provides a behavioural influence on driving, the constant impact of it will act as an emergence of optimal driving behaviour amongst the road users and benefit the environmental safety as well.

The task of educating or training through a gamified environment aims to simplify the learning process by adding a factor of amusement. Gamifying mechanism provides an interaction and engagement of users with the gameplay and this could motivate the users in achieving optimal outcomes. The process of gamifying a non-game context, provides a better learning possibilities for the users [43]. Usually the traditional games focus on the engagement of the users with the gameplay by delivering a good amount of interactivity and various properties on the game scene for entertainment. The serious games focus on providing information and training on some context or to instil certain concepts through a gameplay [44] [45]. Designing a serious game is highly challenging as it should seal the information and entertainment in the same box and required to be developed in a way that it is practically usable over the long run for the continuous development of the user. The core of serious game should be equally balanced between the information and entertainment, if there is an imbalance in this, then the informative content becomes trivial.

Usage of an instructional design (application of technology and multimedia) in crafting a serious game framework can be a significant aspect of a serious game design [46]. The incorporation of instructional design and serious games can enable higher level of learning outcomes, as the design will involve necessary game characteristics (competitions, goals, challenges and etc.) and learning objectives[47].With all these assets, serious games can be exploited in automotive domain to foster green driving (maintaining optimal driving behaviour constantly, without rapid harsh events). The driver behaviour has a major contribution in road safety and green mobility aspects. When analysing the driver behaviour, it’s important to understand the difference between the driver behaviour and the driver performance, where - the driver performance is referred as the abilities (such as skills, knowledge, and cognitive

(17)

GAUTAM RAVINDRA DANGE 17

abilities) of the driver and the driver behaviour is the preference of driver from the experiences gained [48]. It’s also a mandatory fact to concern about the safety, while developing applications that would captivate driving behaviour. These applications should also consider the aspects like driver distraction, immersion on secondary tasks and etc., which could grab driver’s attention. Because, the involvement of drivers in secondary tasks while driving can cause on-road hazards [49]. So, one highly recommended target was to enrich the knowledge and awareness of the driver by an infotainment system, which manages to gamify the driver performance. I propose a gamified environment for drivers to promote green and safe driving, where drivers will not only improve their own performance but also can give a contribution for better driving practices of others in their region by designing their own competition with Green Driving or safe driving criteria’s and by announcing some rewards. The framework allows this to happen in few steps and users need minimal actions to participate into these competitions. The game environment allows the player to visualize the performance on the game screen without direct user interaction to the game. This provides immediate feedback of the performance on game environment and the process of immediate feedback would cultivate a procedural learning [50] and enables user engagement [51] [52]. The immediate feedback also provides an understanding to the user about the driving performance. Especially, when driving behaviour is bad with more harsh driving events (such as: harsh brake, high acceleration and etc.,), the user can visualize a downfall in the game. Thus the impacts of driving pattern on gameplay helps the users to understand driving performance and improvise the driving behaviour. Incentivizing the user performance provides a motivation for the users to progress and improvise the driving traits from the gameplay [53].

Road safety and green mobility are two major concerns for making transportation efficient and convenient for road users. The increase of vehicles and transportation needs of users have created problems such as traffic congestion, accidents and environmental pollution in major cities [55] [56] and these problems remain as hindrance for green and safe mobility of users. While inspecting the issues for the cause of on-road hazards, driver behaviour has the maximum contribution [57] [58]. Driver behaviour not only accounts to fatalities but also in fuel consumption and environmental pollution [59]. As the driving controls (managing speed, braking and gear change) depend on the driver, it causes high fuel consumption when these parameters are not used properly [60].The fuel used during the trip is associated with the emission of pollutant gases, and the driving style of driver determines the vehicle’s quality in the longer run [61][62]. The methods like eco-driving [63] and eco-routing [64] are deployed for saving fuel

(18)

GAUTAM RAVINDRA DANGE 18

and improvising driving style. Driver behaviour assessment through in-car models are getting more prevalent, and in-car driver support models have a direct association with drivers. These models are used to assist drivers in various tasks involving contextual, behavioural and for the betterment of performance as well. The driver support systems have more tendency to eliminate the on-road impacts and fatalities caused due to driver behaviour [65]. For, the purpose of estimating driver performance, the approaches involving the extraction of vehicle signals through CAN (Controller Area Network), or smartphone sensors are used [66] [67] and later these vehicle signals are associated with driver performance. The car manufacturers have taken smart driving initiatives, which enables driver to track and share driving behaviour with peers. There brings better awareness in drivers about their driving. Carwings of Nissan is one such application, which represents the fuel consumption status, money spent vs. buying gas and a comparison of performance with peers [68]. Car2Go, a car-sharing service promotes gamification by grading drivers for environment-friendly driving habits based on acceleration and braking behaviours [69]. From the state-of-the-art analysis, there are numerous approaches, which are addressing the issue of road safety factors from the viewpoint of driver behaviour. By providing necessary knowledge about the driving pattern, a training aspect can be induced in drivers and eventually training of drivers can improve green driving behaviour [70]. The use of gamification process for improvising driver behaviour can benefit the road users in rectifying harsh manoeuvres, understanding performance outcomes and maintaining optimal driving behaviour consistently. Taking all these aspects in consideration, will discuss the vehicle signal processing and real-time gamification models for enhancing the driver behaviour.

The research and experimentation mentioned in the references are either of simulation or driver specific score for himself or for insurance company. Few of the experiments are performed as demonstration tests in specific environment but not having any scope for day-to-day use by the vehicle drivers. The objective of this thesis is to bridge the gap of serious games experimentation in transportation sector and its real-time use in daily life. Framework uncovers experimented methodologies including evaluation of driver performance, Automation of those evaluations by reducing an interaction of vehicle drivers and plug-ins availability for the new games by leveraging the framework. Framework has features like Comparative scores of the drivers driving on the same routes, their historical performance of rides and Application dashboard with details of winner, awards and its graphical data representation. These features would bring entertainment, knowledge and awareness under the same hood. The mention features comparison of drivers’ score is dependent of many factors, which includes location,

(19)

GAUTAM RAVINDRA DANGE 19

time, distance of travel, exact route of travel, time required to complete the travel and how many users are in the same competition. When these factors are satisfied then only Green driving or safe driving criteria’s can be applied for the comparison of their scores. To make this possible framework has server services which keeps tract of competition routes, time slots of competitions, routes of competition, users subscriptions for the competitions and their temporary scores on specific intervals. Input for the serious game is evaluation of the driver performance, so evaluation should be correct and thesis provides detail of scenarios, concerns and findings. There are still many barriers which includes comparison of driver performance of different car models, comparing smart phone based evaluation with CAN bus based evaluation, Automatic start and stop of evaluation with minimal or no configurations by the vehicle driver.

(20)

GAUTAM RAVINDRA DANGE 20

3. Assumptions and Scope

The Evaluators implemented are having few assumptions about the safety, Green driving and their trade-offs against each other in certain conditions. In practical demonstrations, vehicle partners were more interested in safe driving over green driving. Although one algorithm can evaluate the safety and green driving at the same time, I had kept split of 40% green and 60% safe drive evaluation in the car bundles. So, if user get 70/100 in green driving and 80/100 in safe driving then user gets 28 + 48= 76/100 in overall. This scoring split is just a preference I had chosen on the day of demonstration, and in same way, when user creates a new competition, this distribution of safety against green driving would be the preference of the competition designer. The User can get this details in competition description and can decide whether to subscribe for this competition or can find alternative one according to this daily route and competition criteria he/she wish to get evaluated in the drive.

Few environmental factors like road condition, rainy season and snowy roads are outside the coverage of this thesis. Few gamifications developed by lab team using this framework and few user studies have got conducted on those, but discussion of user test is also out of coverage of this thesis.

There are 3 smartphone applications have got used in this framework. 1) In-Application evaluator 2) SG-CB app 3) Automatic Staged Competition Application. The SG-CB Application is the only one application which is mandatory to be used by the vehicle driver. This Application does not serve any purpose when driver is driving, and it required to be used when driver is not driving. So this application won’t cause in distraction to the driver. Another application is Automatic Staged Competition Application, which is used to create competitions and subscribe to competitions by other users. So, this is also used very often and does not required to be used when driving. There is In-Application evaluator which in-turn has a relation with driving because smartphone is required to be mounted in the front desk of the car, with application instance running. When this application runs, values from phone sensors gets processed and the evaluation of the driver happens continuously and results sent to the server on regular interval. Considering the fact that user does not required to interact with the application after he start the application, distraction is almost negligible.

(21)

GAUTAM RAVINDRA DANGE 21

4.

Architecture

The major tasks of this approach involves the evaluation of harsh patterns from the acquired vehicle signals (Speed, RPM, acceleration, etc.,), understanding the driver behaviour and to enhance the better driving traits. The vehicle comprises of an OSGi environment [7], where the evaluators and signal receivers are deployed in the OSGi environment (knopflerfish) as bundles (developed in java, JAR bundles). Vehicle Signals are transmitted over to VDP consumer bundle using VDP APIs in the OSGi framework and from which the signals are assessed by driver performance assessment module (comprises of the three evaluators) and then the evaluation scores are sent to the cloud server using RESTful API by HTTP request-response. The data from the server is accessed using a smartphone application called SG-CB (Serious games and community building) for Graphical representation, statistical reports, and for Serious Games development.

(22)

GAUTAM RAVINDRA DANGE 22

Figure 1: Overview of Framework Architecture and Communication

As fig. 1 and fig. 2 shows, the car has 2 evaluators, In-car evaluation and smartphone-based evaluation also called as In-Application evaluation. The results are getting sent to cloud server and few gamifications are getting reflected according to the scores of the driver. The Game logic is a gamification, which have developed by university of Genova, team members and also it is possible for future research to use this platform and create new games.

(23)

GAUTAM RAVINDRA DANGE 23

Figure 3: Data flow from Evaluation to Visualization and Gamification

As explained in fig. 3, the client modules are Evaluators, SG-SB application, Automatic competition configuration Application and Games. The server is responsible for managing the communication using HTTP request response calls. The cloud has Apache Tomcat server running with RESTful API implementation and MySQL database. The Evaluation, Games and SG-CB application communicate with server using HTTP request response requests. The server is having different modules to serve different purposes.

In-car evaluator uses CAN bus signals from the car and using algorithms process these signals to evaluate user score, detects harsh events. The In-Application evaluator uses mobile sensors like Gyroscope, Accelerometer and GPS location for evaluating the vehicle driver. The In-car evaluator uses different algorithms to analyse the data and evaluate the user score in real time. These algorithms will be explained in next session. The In-car evaluator and In-Application evaluators communicate the scores, events, times, GPS location with Longitude and latitude and user information to the server using HTTP request (refer fig. 4). The server responds with insertion confirmation.

The Automatic Competition Configuration Manager is responsible to support automatic configuration of evaluation and setting up competition environment for the user. This provides ease of user for vehicle drivers for competing in Green Driver competitions just by subscribing a competition and having evaluator turned on for the evaluation. More about this module will be explained in session 6.

(24)

GAUTAM RAVINDRA DANGE 24

Figure 4: HTTP request-response based RESTful APIs

The results of users send to server by evaluators are accessible to SG-SB application for the presentation to the user about his/her performance, which includes performance in particular competition, historical performance over the time and performance comparison with other users. Users also can check their harsh manoeuvres on the road on google map and also can check their prize related information and virtual coins which they achieved in case of winning the competition.

The partial scores of the user on regular interval are accessible on the server for the implementation of Games. The Gamification module is responsible for providing the information of user score on regular time interval. This exposes a new way of segregation in serious games as Game designer can develop a game without being aware of exact methodology of the evaluation.

4.1 Social Gaming prototype scenario

The social gaming scenario consists of various competitions (drive around a location, for example: a city) comprising a peer evaluation, where the competition involves the participation of multiple users and each user in the competition will be evaluated and granted a score out of 100 (scores will be calculated based on the results of the evaluators used). The competition user holding a highest score will be declared as the winner of competition and will be awarded with the virtual coins. The users can get virtual coins at any time by exhibiting safe and green driving behaviour, as virtual coins are continuous evaluation of driver performance independent of peers, routes and competitions.

(25)

GAUTAM RAVINDRA DANGE 25

4.2 Competitions:

In competitions approach, the user can take part in a time framed competition (a competition might last from 10 – 15 minutes depending on the locality) by subscribing for the open competitions (the competitions are opened on timely basis) and exhibit better driving behaviour to surpass the peers in competition. The term competition in this occasion can be defined as a geographical location associated with any road network in a city. In a competition, users are evaluated for their performance by the two green drive evaluators on the basis of vehicle signals (acceleration, brake, engine RPM, and speed) and the smartphone signals (GPS, accelerometer, and gyroscope). While, on the go the users can check the scores of their subscribed competitions and also can look for the fluctuations in scores based on the performance in competition. On completion of a competition, users can check their detailed report of scores, rankings, performances and comparison with peers. The competition strategy, grant scores and generate the rankings based on the comparison of user performance with the performance of peers, so this aspect comprises of impact in evaluation from various users in the competition. Some virtual coins are also granted for scores secured in the competitions. Whilst, analysing the performance in competition, the users are also assessed on harsh driving events such as harsh braking, high acceleration and high levels of engine RPM. At the end of competition, the tracked harsh events are displayed on Google maps along with the harshness level (high or average) and this methodology would serve the purpose of training drivers by making aware of the harsh patterns exhibited during the drive.

4.3 SG-CB (Serious Games & Community Building) Application

HMI has been designed to have a very little impact on the driver, while allowing third parties (e.g., a passenger) to assess to information in real-time. As shown in fig. 5, few performance visualization and gamification interfaces are provided in this application. The historical data of performance and comparisons with other users are available for user. There is separate session for the Performance Visualization and Gamification to discuss more about this in details.

(26)

GAUTAM RAVINDRA DANGE 26

Figure 5: Serious Games and Community Building Application introduction

4.4 Virtual Coins

The user gains virtual coins (rewards granted based on the performance) for optimal driving. The virtual coins are accumulated in the virtual bank and they can be used on real-world entities such as purchasing bus tickets, reservation of parking space and etc. Fig. 6 shows snapshot of user’s virtual coins and using it for purchase of bus tickets. This is most common incentives used in TEAM project, research team in my lab had come up with this concept and for all application linked with the framework, and Virtual coins are supported as one of the medium of incentives.

(27)
(28)

GAUTAM RAVINDRA DANGE 28

5.

Vehicle Performance Evaluation

The gamification aspect is based on green drive performance of driver by assessment of various vehicle signals (acceleration, brake, speed and RPM). The vehicle signals are extracted from CAN bus, and they are sent to an in-car evaluator module (units for signal processing and performance evaluation are deployed as bundles in OSGi platform [71]). In the OSGi environment, the bundles interprets signal values from the CAN Bus and transfers it to VDP (Vehicle Data Provider) bundle in OSGi environment, and subsequently sends the signals to different evaluator applications which uses VPC (Vehicle Data consumer) service, which also runs in OSGi environment and have subscribed for signals. Due to this approach of pipelining, the addition of new evaluator would not require any code changes and configuration change can allow access to that.

The Green Driving and Safe Driving are measurable features of better driving practices. The User driving performance can be determined using vehicle signals. This section will provide explanation about how vehicle data is collected from vehicle, how raw signals are processed to get user driving performance knowledge and how process can be standardized to use it for different vehicle models. Evaluation activity is divided in to two phases, and mentioned in this section as pre-Test and post-Test phase. The pre-Test phase / Data Analysis covers the activities and research methodologies when the actual site tests were not available and models used to run on data sets given my different car models. The Run-time phase/ Run time Evaluation covers the real-time evaluation of the driver when OSGi environment running inside a car and real-time values of signals gets analysed to evaluate a driver performance.

Pre-Test Phase / Data Analysis:

Activity started with the collection of the raw signals data from the Car, by installing log jars in to OSGi car environment. OSGi Jar Bundle, created log files for different car models for 20 minute long car test rides. This different car models data became available in log files. These acquired signal samples from FIAT, BWM and Mercedes Benz analysed to evaluate a vehicle driver performance.

We asked FIAT, BMW and Mercedes Benz car drivers to drive with different patterns which includes: Safe drive, Green drive, Harsh and unsafe drive. The pattern of driving was already known for each drive, and samples log collected from the car for each of those drives. The harsh manoeuvres includes high RPM, harsh brakes, harsh accelerations and brake with

(29)

GAUTAM RAVINDRA DANGE 29

steering wheel rotation. Collected Samples analysed in python to understand nature of different signals. It is importance to analyse the data when vehicle is moving, so, most of the observations presented in this whole section are filtered with for speed >0. These drives has got conducted in training route which has highway equivalent traffic conditions. So described graphs in this section are depicting highway drive collected signals.

This research focus on practical way of evaluating driver performance and its gamification. The framework would allow evaluation developer to plugin new evaluators to the framework and game designers to use this evaluators scores to use as an input for their games. Harsh or rash instances are fraction of single percent of whole driving, so, evaluation need to be penalty based. In the real time competition, whenever driver apply any harsh pattern, Algorithm detects this moves and assign a Penalty, for this pattern. This penalty will inversely affect this competition score. If Harshness of driving is significantly high, and driver get high penalty, then driver need to be notified about this behaviour with the details about date, time, harness pattern and location where this harsh behaviour is observed.

Designed and implemented algorithms for estimating a driver performance are deployed on OSGi environment as JAR bundles. Evaluation patterns provide more scope for understanding driver performance on individual samples and accumulated event windows. The extracted vehicle signals are mapped and processed for understanding the driving behavior; harsh patterns from the signals are spotted and penalized. Harsh patterns comprise rapid brakes, frequent high acceleration, and the combination of steering wheel angle and brakes

5.1 Vehicle Data Analysis:

This section provides vehicle signals analysed using Jupyter notebook in Python. As the graphs explain, there are many different signals from the vehicle. By looking at the graphs, it is possible to get an idea of the normal range of every signal. Signal analysis have got performed in python, for understanding the behaviour of vehicle signals on different manoeuvres by the vehicle driver. To better evaluation the signals where speed is 0 are removed, that is vehicle is parked, or waiting on the signal. These analysis gives information of maximum possible signal values as well as average and standard deviations in their values.

The collected signal sample sets were used to perform some data analysis to understand the relationship between car signal and their relation with driver’s manoeuvres. I was knowing in prior about the driver manoeuvres, as these rides were designed to be harsh and smooth, in advance. So, the known behaviour and the signal values were required to be mapped and

(30)

GAUTAM RAVINDRA DANGE 30

crosschecked against each other. Although the information of event is known, the event like brake actually takes multiple samples and few of them can be a good braking subset and few values with harsh values. So, to analyse the harshness knowing the natural pattern of the signals is recommended. The Data analysis phase has 3 sample data sets and fig. 7, fig. 8 and fig. 9 shows BMW, FIAT and Mercedes-Benz car signals snapshot respectively. These graphs gives an idea about the nature of signal, and few correlation graphs provide more information about signal relation with another signals.

Figure 7: BMW Signals from CAN bus collected log

BMW Nissan snapshot represent few samples of the data set, few observation about BMW signal samples :

1. Longitudinal Acceleration Range :

Approximately -5 yo +5 gravity force m/s2. As the whole data set -6 yo +5.5

2. Lateral Accelearation Range : Approximately -2 to +4. Approximately -3.8to +5.

(31)

GAUTAM RAVINDRA DANGE 31

3. Brake values range :

Approximately : 0 to -5000 As the whole data set 0 to -6100 4. RPM range :

Approximately : 0 to 6500 As the whole data set 0 to 7100

Figure 8: FIAT car Signals from CAN bus collected log

FIAT 500 car snapshot represent few samples of the data set, few observation about FIAT 500 signal samples:

1. Longitudinal Acceleration Range :

Approximately -4 yo +2 gravity force m/s2. As the whole data set -4.2 yo +3.5

2. Lateral Accelearation Range : Approximately -2 to +2. Approximately -2.2 to 2.3. 3. Brake values range :

Approximately : 0 to 30 As the whole data set 0 to 60 4. RPM range :

(32)

GAUTAM RAVINDRA DANGE 32

Approximately : 0 to 4500 As the whole data set 0 to 5200

Figure 9: Mercedes-Benz car Signals from CAN bus collected log

Mercedes Benz car snapshot represent few samples of the data set, few observation about Mercedes Benz signal samples:

1. Acceleration Range :

0 to 100 gravity force m/s2. 2. Brake values range :

Approximately : 0 to 2500 As the whole data set 0 to 4200 3. RPM range :

Approximately : 0 to 5000 As the whole data set 0 to 5200

(33)

GAUTAM RAVINDRA DANGE 33

5.1.1. Basis of Evaluation:

The driver evaluation is based on 9 signals: Acceleration, Longitudinal Acceleration, Latitudinal Acceleration, Brake, Speed, RPM, Steering wheel angle, steering wheel rotation speed, and GPS latitude and longitude values. The pattern of braking signal itself provide vital information of drivers driving pattern. Observation shows relation between Speed and Brake harsh moves, as if brake is applied harshly when vehicle speed is low, it will not end up in unsafe condition but risk will be considered higher as speed is higher. The so, various dimensions come in to picture to evaluate penalty. Another important factor is duration of brake or number of multiple brakes in specific period, which affect the scores of driver.

5.1.2 Normalizations

As it can be observed from the fig. 10, fig. 11 and fig. 12 the brake, RPM, and acceleration values are different for different car models, and that rises a concern for the algorithm to mine the data correctly. The scaling the data with normalization or standardization were the options. As fig. 7 shows the BMW treat the Brake signal in negative number, whereas FIAT and Mercedes Benz car has positive high value for the brake. So, decision was to go with normalization and use the absolute values for the signals, which can resolve problem with negative values. Considering the magnitude value away from the 0 shows harsh event, the absolute value usage and normalization does not create a situation where loss of information can be possible. The following fig. 10, fig.11 and fig.12 shows the normalization of BMW, FIAT and Mercedes-Benz car signals respectively.

(34)

GAUTAM RAVINDRA DANGE 34

Figure 10: BMW car Normalization signals

The fig. 10 shows the normalized values for the BMW Nissan Car model signal. The BMW Nissan car had negative values for the brake ranging from 0 to -5000, now these values are in range of 0 to positive 1, and the nature of brake signal is unaltered. As well as, the Longitudinal Acceleration and Lateral Acceleration also have only positive values. As the magnitude of the value only show the harshness, converting these from negative to positive does not results in to loss of information. There is another 1 signal is introduced which is a Brake Vs speed signal. As brake value with low speed does not have a great impact on harshness, but brake value with high speed values results in to harsh brake event. So, this signal have added to visually

(35)

GAUTAM RAVINDRA DANGE 35

understand such cases in the test ride. Having these values in a range of 0 and 1, gives a better chance of having 1 algorithm which can evaluate the vehicle driver performances on different car models.

Figure 11: FIAT car Normalization signals

The fig. 11 shows the normalized values for the FIAT 500 car signals. The Longitudinal Acceleration and Lateral Acceleration values are not only in positive range. Same as per BMW, new signal Brake vs speed was introduced to present the harsh event graphically.

(36)

GAUTAM RAVINDRA DANGE 36

The fig. 12 shows the normalized values for Mercedes-benz s-class car signals. Although all the signals’ values of Mercedes-benz are positive, only normalization was required to perform for achieving value range from 0 to 1.

Figure 12: Mercedes-Benz car Normalization signals

5.1.3 Observations, Possible False conceptions and clarifications:

The normalization graphs of all the car can give an idea about the signal nature, but also create some false conception. For example, by observing Mercedes-Benzes signal for brake vs speed, it seems evident that the driver has applied lot of harsh brakes and he/she was aggressive on the road. But, unless until have a knowledge of the real optimal range of brake, it is not possible to conclude if behaviour was harsh or normal. To solve this problem, I have used

(37)

GAUTAM RAVINDRA DANGE 37

average and standard deviations and implemented them in program with dynamic update. The value reflect whenever there is a new value for maximum, minimum, average of last n-samples and standard deviation in last n-samples.

5.2 Signal Correlations:

For better understanding of the relative changes in signals, I have correlated those signals. The Graphs below shows the correlations between those signals.

Figure 13: Correlation of Longitudinal (g-force m/s2) and Lateral Acceleration (g-force m/s2)

The fig. 13 shows the relation between lateral and longitudinal acceleration. It can be observed that, either of these value generally can deviate from 0, but both values deviating from 0 with high margin are relatively very less.

(38)

GAUTAM RAVINDRA DANGE 38

Figure 14: Correlation of Lateral Acceleration (g-force m/s2) of Brake torque (Newton Meter Nm)

Fig. 14 shows the relationship between the Brake and Lateral Acceleration. It can be observed that when value for brake is high, lateral acceleration value is deviating toward -5, and it means driver is preparing for the turn (specifically right turn or stopping for the parking).

Figure 15: Correlation of Lateral Acceleration (g-force m/s2) and RPM

The fig. 15 shows the relationship between RPM and Lateral Acceleration, it can be observed that with high value of RPM, Lateral acceleration is deviating towards positive higher values, and it means driver is taking left turn and can infer that vehicle speed is high.

(39)

GAUTAM RAVINDRA DANGE 39

Figure 16: Correlation of Lateral Acceleration (g-force m/s2) and Speed (km/h)

The fig. 16 represent the relation of Speed signal to Lateral Acceleration and basically there is no relation between Lateral Acceleration and speed could be found in this snapshot, as with speed values all variations of lateral acceleration values spread approximately in same density.

Figure 17: Correlation of Longitudinal Acceleration (g-force m/s2) and Brake torque (Newton Meter Nm)

As longitudinal acceleration fig. 17 represents the acceleration and deceleration movement of the car, graph depicts that the higher value of brake vehicle movement further or reverse is getting restricted.

(40)

GAUTAM RAVINDRA DANGE 40

Figure 18: Correlation of Longitudinal Acceleration (g-force m/s2) and RPM

The fig. 18 represents the correlation between RPM and Longitudinal acceleration, and it can be observed that, with high RPM car has achieved smooth speed, whereas with low PRM vehicle speed has an ever-changing value unless speed is around 0.

Figure 19: Correlation of Longitudinal Acceleration (g-force m/s2) and Speed (km/h)

Fig. 19(a) represent the relationship between Speed and Longitudinal acceleration. For better understanding between these 2 signals, 2 snapshots are represented. It can be observed in first fig. 19 (a) that, with high speed value, Acceleration longitudinal values deviate towards positive(acceleration), whereas with small speed values it deviate towards negative

(41)

GAUTAM RAVINDRA DANGE 41

(deceleration). In fig. 19 (b) it can be observed that, the speed value around 50, has minimal longitudinal acceleration. So, there is some new information which can’t be revealed only by relationship of signals.

Figure 20: Correlation of Brake torque (Newton Meter Nm) and RPM

The fig. 20 shows the straight forward relationship between RPM and Brake. As brake goes higher vehicle goes to speed 0 so resulting low RPM. If brake is low it allows vehicle to speed up and RPM can be high according to vehicle speed, gear and road condition.

(42)

GAUTAM RAVINDRA DANGE 42

The brake and speed relationship is depicted in a fig. 21. If brake value is high then speed value is expected to be low, whereas for having high speed, brake value should be low. Although these are snapshots of all signals, so it is possible to find few samples with high brake with high speed. This is because speed starts getting impacted for next subsequent samples. So if we plot brake and speed signals against time axis on x (as shown in fig. 10, 11, and 12), it can be observed that with high brake values vehicle speed gets changes inversely.

This information is helpful in understanding of brake, as with high speed even medium brake value can have high impact, and in normal cases this situation is rare. So, for penalizing user with reduction in score, the brake and speed relation plays important role.

Figure 22: Correlation of RPM and Speed (km/h)

The fig. 22 shows the RPM and Speed relation, the diagonal lines actually represent a gears of vehicle, the RPM is expected to go higher with speed, but with low speed RPM going very high is not recommended. This gives a generic idea of operating RPM at particular speed level. Although this observation is not sufficient to find out if driving is green, but gives RPM information comparing with average RPM value of same speed range.

5.3 Penalty Evaluation

There are few more contributors in Penalty calculation which includes, high engine RPM, Harsh Acceleration, high Longitudinal Acceleration and High Latitudinal Acceleration.

(43)

GAUTAM RAVINDRA DANGE 43

In fig. 23, “Original + above thr = 500”, This snapshot has been collected considering threshold is 500 for the Brake, and the above threshold events are marked in red whereas below threshold brake values are in blue. In “Penalty Event – PE”, the binary value of event window, if it is in penalty is displayed. The “Number of PE in last 30 samples”, shows the penalty calculation of last 30 samples. “Energy of PE – Area under the curve” shows the calculation of total penalty in the window which will be added together for final penalty calculation for the Brake signal. The “Even + intensity” shows the calculated penalty value in specific penalty window. It can be observed that around 380 to 430 samples, brake values were harsh with high speed, which drastically has reduced speed from 90km/h around to 40km/h. Accordingly its penalty calculation is showing number around 0.9 which is in fact very high penalty, and thus score of the user will get affected inversely. There are also few more instances of brake with some moderately harsh values and thus they are not having high penalty.

(44)

GAUTAM RAVINDRA DANGE 44

Figure 23: Brake Penalty calculation against speed with sliding window

In fig. 24, figure represent another case of brake, the “Original + above thr = 500”, considering threshold is 500 for the Brake, and the above threshold events are marked in red whereas below threshold brake values are in blue. In “Penalty Event – PE”, the binary value of event window, if it is in penalty is displayed. The “Number of PE in last 30 samples”, shows the penalty calculation of last 30 samples. “Energy of PE – Area under the curve” shows the calculation of total penalty in the window which will be added together for final penalty calculation for the Brake signal. The “Even + intensity” shows the calculated penalty value in specific penalty window. It can be observed that around 120, 270, 780 920, brake values were harsh with high speed, which drastically has reduced speed from 90km/h around to 40km/h. Accordingly its penalty calculation is showing number around 0.8 & 0.9 which is in fact very high

Riferimenti

Documenti correlati

Section 4 presents an abstract model based on continuous time Markov chains and Stochastic Automata for synchronising par- allel reversible computations.. We discuss the application

The index is used to calculate the technical efficiency of 500 high-performing computing systems listed in the TOP500 supercomputers database.. Finally,

When the database of interest has been created, the user will generate the map of all the possible linear peptides containing one and two cysteines for each protein and combine them

the seminar to be sent to the various Salesian Bulletins and Provincial Neivslbtters, which could publish the same with due modifications and/or after translation, Fr

Based on the analysis of unfavorable atmospheric conditions for propagation of radio waves, it can be noticed that the impact on the controllability of the UAV

In the next stage of investigations there were determined mathematical models with the concentrate recovery, concentrate grade, yield of concentrate and tails grade as the

3 Abaco degli elementi strutturali modellati con BIM dello stato di fatto della pista, il risul- tato è stato ottenuto dalla rielaborazione delle geometrie ottenute dal

[15] ATLAS Collaboration, A search for new physics in dijet mass and angular distributions in pp collisions at p ffiffiffi s ¼ 7 TeV measured with the ATLAS detector, New J.. High